Land-Cover Change Detection: Measuring What Moved Between Two Dates
A single map tells you what is there; two maps, compared properly, tell you what changed — where forest became farmland, where a city crept outward. Change detection is the discipline of extracting that difference reliably.
What it actually measures
Two broad families exist. Post-classification comparison overlays two classified maps and reads a from–to transition matrix; spectral differencing subtracts image values (or indices) between dates and thresholds the change. The first gives labelled transitions, the second gives magnitude and catches subtler change.
The math behind it
How to interpret the output
Change vs no-change is set by a threshold or a from–to class transition matrix.
What it runs on
| Operates on | Two classified maps or images from different dates — not tied to specific spectral bands. |
Real-world use
It quantifies deforestation, urban expansion, agricultural conversion, wetland loss and post-disaster damage — anywhere the question is how much, and where, the land changed.
Watch out for
Its biggest enemy is false change from mis-registration and from radiometric or seasonal differences between dates. Co-register precisely, normalise or atmospherically correct both images, and use anniversary dates to hold phenology constant.
Compute Change Detection on your own study area
Skip the code. Draw or upload a boundary and Spatial Research Suite runs Change Detection on live data — with cloud masking, exports and citations built in.
Run this analysis in GISforus →Frequently asked
What is the difference between post-classification and image-differencing change detection?
Post-classification compares two thematic maps and reports from–to transitions; image differencing subtracts pixel or index values and thresholds the magnitude of change.
What causes false change?
Mis-registration, and radiometric or seasonal differences between the two dates — all of which can look like real change unless corrected.